inference/research
verifiedhigh · 0.88updated 2026-10-03

MicroMix: Efficient Mixed-Precision Quantization with Microscaling Formats for Large Language Models

paper

Fields

id
sources/qlab-micromix-paper
URL
https://arxiv.org/abs/2508.02343
Publisher
arXiv
Kind
paper
Published
2025-08-04
Accessed
2026-10-03
notes
Blackwell-targeted mixed MXFP4/MXFP6/MXFP8 kernel; source for the MXFP6 record.

Notes

Blackwell-targeted mixed MXFP4/MXFP6/MXFP8 kernel; source for the MXFP6 record.

Cross-references

This record references no other records.